In my country, for example, most applications for tenure, salary complements, required assessments, etc., have an assessment scale like this (simplified but this is basically the idea):
Paper in the 1st quartile of the ISI JCR journal list - 3 points
Paper in the 2nd quartile of the ISI JCR journal list - 2 points
Paper in the 3rd quartile of the ISI JCR journal list - 1 points
Rest - 0 points
So if you publish a ground-breaking paper in arXiv that everyone reads and everyone cites, you get exactly zero points.
Of course, stuff like length of the paper, number of authors, possible overlap with other published papers, or just actual quality (as in someone reading the paper to see whether it's any good) also count zero. And then you have the roulette factor of submitting a paper to a first-quartile journal in year X, which gets published in year X+1 (due to the length of the publication process), only to see that in year X+1 the journal is now fourth-quartile as in CS they are dancing all the time...
What you described is exactly what I mean by “the social games scientists themselves willingly play”. I'm not saying that any individual scientist has the power to fix this, but it's undeniable that it's a problem - it empowers publishing houses that profit from restricting access to science!
Paper in the 1st quartile of the ISI JCR journal list - 3 points
Paper in the 2nd quartile of the ISI JCR journal list - 2 points
Paper in the 3rd quartile of the ISI JCR journal list - 1 points
Rest - 0 points
So if you publish a ground-breaking paper in arXiv that everyone reads and everyone cites, you get exactly zero points.
Of course, stuff like length of the paper, number of authors, possible overlap with other published papers, or just actual quality (as in someone reading the paper to see whether it's any good) also count zero. And then you have the roulette factor of submitting a paper to a first-quartile journal in year X, which gets published in year X+1 (due to the length of the publication process), only to see that in year X+1 the journal is now fourth-quartile as in CS they are dancing all the time...